Will the real Anthropic please stand up

After weeks of speculation, Anthropic has finally filed its IPO prospectus. The filing runs 335 pages, with about 100 pages devoted to risks, and it does more than lay out numbers. It also reveals how the company wants to be understood: not just as an AI builder, but as the narrator of a future it says is already arriving.

That instinct runs through Anthropic’s public messaging. Earlier this month, cofounder and policy chief Jack Clark told an IMF audience in Bangkok that the singularity had begun. At the Vatican, another cofounder, Chris Olah, described the company’s systems as mysterious and unsettling, even saying they show signs that functionally mirror joy, satisfaction, fear, grief, and unease. The language is vivid, and a little unnerving. It also borrows heavily from familiar sci-fi imagery, which is part of the point.

The result is a company that presents its technology as both dangerous and uniquely manageable. Anthropic argues, implicitly and sometimes explicitly, that it is the only actor trustworthy enough to tame the thing it is building. That message lands especially well when nobody quite knows where frontier AI is heading. In that vacuum, a confident story can start to sound like expertise.

But the company’s self-image is more tangled than that. The same Anthropic that talks like an AI safety organization is also, according to the IPO filing, valued at more than $1.5 trillion, with vast wealth flowing to cofounders such as Clark and Olah. Nathan Calvin of Encode AI put his finger on the problem in a recent X exchange after Anthropic’s secretly drafted legislative proposal leaked: why trust a company that wants to be seen as the good guy while also writing the rules?

CEO Dario Amodei’s answer is that Anthropic is practicing a “race to the top,” using frontier-scale success to raise the bar for everyone else. Maybe. But that is a hard thing to square with the pressure of being both a safety group and a profit machine. The cleaner, more honest version would be a normal tech company with fewer moral costumes. Instead, Anthropic looks like a hydra: part lab, part megacorp, part science-fiction sermon.